Gait Based Biometric Authentication System with Reduced Search Space

Abstract

Deployment of human gait in developing new tools for security enhancement has received growing attention in modern era. Since the efficiency of any algorithm depends on the size of search space, the aim is to propose a novel approach to reduce the search space. In order to achieve this, the database is split into two based on gender and the search is restricted in the identified gender database. Then highly discriminant gait features are selected by forward sequential feature selection algorithm in the confined space. Experimental results evaluated on the benchmark CASIA B gait dataset with the newly proposed combined classifier kNN-SVM, shows less False Acceptance Rate (FAR) and less False Rejection Rate (FRR).